#+title: Simulated Sense of Hearing

 #+author: Robert McIntyre

 #+email: rlm@mit.edu

 #+description: Simulating multiple listeners and the sense of hearing in jMonkeyEngine3

 #+keywords: simulated hearing, openal, clojure, jMonkeyEngine3, LWJGL, AI

 #+SETUPFILE: ../../aurellem/org/setup.org

 #+INCLUDE: ../../aurellem/org/level-0.org

 #+BABEL: :exports both :noweb yes :cache no :mkdirp yes

 

 * Hearing

 

 At the end of this post I will have simulated ears that work the same

 way as the simulated eyes in the last post.  I will be able to place

 any number of ear-nodes in a blender file, and they will bind to the

 closest physical object and follow it as it moves around. Each ear

 will provide access to the sound data it picks up between every frame.

 

 Hearing is one of the more difficult senses to simulate, because there

 is less support for obtaining the actual sound data that is processed

 by jMonkeyEngine3. There is no "split-screen" support for rendering

 sound from different points of view, and there is no way to directly

 access the rendered sound data.

 

 ** Brief Description of jMonkeyEngine's Sound System

 

 jMonkeyEngine's sound system works as follows:

 

  - jMonkeyEngine uses the =AppSettings= for the particular application

    to determine what sort of =AudioRenderer= should be used.

  - Although some support is provided for multiple AudioRendering

    backends, jMonkeyEngine at the time of this writing will either

    pick no =AudioRenderer= at all, or the =LwjglAudioRenderer=.

  - jMonkeyEngine tries to figure out what sort of system you're

    running and extracts the appropriate native libraries.

  - The =LwjglAudioRenderer= uses the [[http://lwjgl.org/][=LWJGL=]] (LightWeight Java Game

    Library) bindings to interface with a C library called [[http://kcat.strangesoft.net/openal.html][=OpenAL=]]

  - =OpenAL= renders the 3D sound and feeds the rendered sound directly

    to any of various sound output devices with which it knows how to

    communicate.

 

 A consequence of this is that there's no way to access the actual

 sound data produced by =OpenAL=. Even worse, =OpenAL= only supports

 one /listener/ (it renders sound data from only one perspective),

 which normally isn't a problem for games, but becomes a problem when

 trying to make multiple AI creatures that can each hear the world from

 a different perspective.

 

 To make many AI creatures in jMonkeyEngine that can each hear the

 world from their own perspective, or to make a single creature with

 many ears, it is necessary to go all the way back to =OpenAL= and

 implement support for simulated hearing there.

 

 * Extending =OpenAL=

 ** =OpenAL= Devices

    

 =OpenAL= goes to great lengths to support many different systems, all

 with different sound capabilities and interfaces.  It accomplishes this

 difficult task by providing code for many different sound backends in

 pseudo-objects called /Devices/.  There's a device for the Linux Open

 Sound System and the Advanced Linux Sound Architecture, there's one

 for Direct Sound on Windows, there's even one for Solaris. =OpenAL=

 solves the problem of platform independence by providing all these

 Devices.

 

 Wrapper libraries such as LWJGL are free to examine the system on

 which they are running and then select an appropriate device for that

 system. 

 

 There are also a few "special" devices that don't interface with any

 particular system.  These include the Null Device, which doesn't do

 anything, and the Wave Device, which writes whatever sound it receives

 to a file, if everything has been set up correctly when configuring

 =OpenAL=.

 

 Actual mixing of the sound data happens in the Devices, and they are

 the only point in the sound rendering process where this data is

 available. 

 

 Therefore, in order to support multiple listeners, and get the sound

 data in a form that the AIs can use, it is necessary to create a new

 Device which supports this features.

 

 ** The Send Device

 Adding a device to OpenAL is rather tricky -- there are five separate

 files in the =OpenAL= source tree that must be modified to do so. I've

 documented this process [[../../audio-send/html/add-new-device.html][here]] for anyone who is interested.

 

 Again, my objectives are:

 

  - Support Multiple Listeners from jMonkeyEngine3

  - Get access to the rendered sound data for further processing from

    clojure.

 

 I named it the "Multiple Audio Send" Deives, or =Send= Device for

 short, since it sends audio data back to the callig application like

 an Aux-Send cable on a mixing board.

 

 Onward to the actual Device!

 

 ** =send.c=

 

 ** Header

 #+name: send-header

 #+begin_src C

 #include "config.h"

 #include <stdlib.h>

 #include "alMain.h"

 #include "AL/al.h"

 #include "AL/alc.h"

 #include "alSource.h"

 #include <jni.h>

 

 //////////////////// Summary

 

 struct send_data;

 struct context_data;

 

 static void addContext(ALCdevice *, ALCcontext *);

 static void syncContexts(ALCcontext *master, ALCcontext *slave);

 static void syncSources(ALsource *master, ALsource *slave, 

 			ALCcontext *masterCtx, ALCcontext *slaveCtx);

 

 static void syncSourcei(ALuint master, ALuint slave,

 			ALCcontext *masterCtx, ALCcontext *ctx2, ALenum param);

 static void syncSourcef(ALuint master, ALuint slave,

 			ALCcontext *masterCtx, ALCcontext *ctx2, ALenum param);

 static void syncSource3f(ALuint master, ALuint slave,

 			ALCcontext *masterCtx, ALCcontext *ctx2, ALenum param);

 

 static void swapInContext(ALCdevice *, struct context_data *);

 static void saveContext(ALCdevice *, struct context_data *);

 static void limitContext(ALCdevice *, ALCcontext *);

 static void unLimitContext(ALCdevice *);

 

 static void init(ALCdevice *);

 static void renderData(ALCdevice *, int samples);

 

 #define UNUSED(x)  (void)(x)

 #+end_src

 

 The main idea behind the Send device is to take advantage of the fact

 that LWJGL only manages one /context/ when using OpenAL.  A /context/

 is like a container that holds samples and keeps track of where the

 listener is.  In order to support multiple listeners, the Send device

 identifies the LWJGL context as the master context, and creates any

 number of slave contexts to represent additional listeners.  Every

 time the device renders sound, it synchronizes every source from the

 master LWJGL context to the slave contexts.  Then, it renders each

 context separately, using a different listener for each one.  The

 rendered sound is made available via JNI to jMonkeyEngine.

 

 To recap, the process is: 

  - Set the LWJGL context as "master" in the =init()= method.

  - Create any number of additional contexts via =addContext()=

  - At every call to =renderData()= sync the master context with the

    slave contexts with =syncContexts()=

  - =syncContexts()= calls =syncSources()= to sync all the sources

    which are in the master context.

  - =limitContext()= and =unLimitContext()= make it possible to render

    only one context at a time.

 

 ** Necessary State

 #+name: send-state

 #+begin_src C

 ////////////////////  State

 

 typedef struct context_data {

   ALfloat ClickRemoval[MAXCHANNELS];

   ALfloat PendingClicks[MAXCHANNELS];

   ALvoid *renderBuffer;

   ALCcontext *ctx;

 } context_data;

 

 typedef struct send_data {

   ALuint size;

   context_data **contexts;

   ALuint numContexts;

   ALuint maxContexts;

 } send_data;

 #+end_src

 

 Switching between contexts is not the normal operation of a Device,

 and one of the problems with doing so is that a Device normally keeps

 around a few pieces of state such as the =ClickRemoval= array above

 which will become corrupted if the contexts are not rendered in

 parallel. The solution is to create a copy of this normally global

 device state for each context, and copy it back and forth into and out

 of the actual device state whenever a context is rendered.

 

 ** Synchronization Macros

 #+name: sync-macros

 #+begin_src C

 ////////////////////  Context Creation / Synchronization

 

 #define _MAKE_SYNC(NAME, INIT_EXPR, GET_EXPR, SET_EXPR)	\

   void NAME (ALuint sourceID1, ALuint sourceID2,	\

 	     ALCcontext *ctx1, ALCcontext *ctx2,	\

 	     ALenum param){				\

     INIT_EXPR;						\

     ALCcontext *current = alcGetCurrentContext();	\

     alcMakeContextCurrent(ctx1);			\

     GET_EXPR;						\

     alcMakeContextCurrent(ctx2);			\

     SET_EXPR;						\

     alcMakeContextCurrent(current);			\

   }

   

 #define MAKE_SYNC(NAME, TYPE, GET, SET)			\

   _MAKE_SYNC(NAME,					\

 	     TYPE value,				\

              GET(sourceID1, param, &value),		\

              SET(sourceID2, param, value))

   

 #define MAKE_SYNC3(NAME, TYPE, GET, SET)				\

   _MAKE_SYNC(NAME,							\

 	     TYPE value1; TYPE value2; TYPE value3;,			\

 	     GET(sourceID1, param, &value1, &value2, &value3),		\

 	     SET(sourceID2, param,  value1,  value2,  value3))

 

 MAKE_SYNC( syncSourcei,  ALint,   alGetSourcei,  alSourcei);

 MAKE_SYNC( syncSourcef,  ALfloat, alGetSourcef,  alSourcef);

 MAKE_SYNC3(syncSource3i, ALint,   alGetSource3i, alSource3i);

 MAKE_SYNC3(syncSource3f, ALfloat, alGetSource3f, alSource3f);

   

 #+end_src

 

 Setting the state of an =OpenAL= source is done with the =alSourcei=,

 =alSourcef=, =alSource3i=, and =alSource3f= functions.  In order to

 completely synchronize two sources, it is necessary to use all of

 them. These macros help to condense the otherwise repetitive

 synchronization code involving these similar low-level =OpenAL= functions.

 

 ** Source Synchronization

 #+name: sync-sources

 #+begin_src C

 void syncSources(ALsource *masterSource, ALsource *slaveSource, 

 		 ALCcontext *masterCtx, ALCcontext *slaveCtx){

   ALuint master = masterSource->source;

   ALuint slave = slaveSource->source;

   ALCcontext *current = alcGetCurrentContext();

 

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_PITCH);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_GAIN);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_MAX_DISTANCE);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_ROLLOFF_FACTOR);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_REFERENCE_DISTANCE);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_MIN_GAIN);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_MAX_GAIN);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_CONE_OUTER_GAIN);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_CONE_INNER_ANGLE);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_CONE_OUTER_ANGLE);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_SEC_OFFSET);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_SAMPLE_OFFSET);

   syncSourcef(master,slave,masterCtx,slaveCtx,AL_BYTE_OFFSET);

     

   syncSource3f(master,slave,masterCtx,slaveCtx,AL_POSITION);

   syncSource3f(master,slave,masterCtx,slaveCtx,AL_VELOCITY);

   syncSource3f(master,slave,masterCtx,slaveCtx,AL_DIRECTION);

   

   syncSourcei(master,slave,masterCtx,slaveCtx,AL_SOURCE_RELATIVE);

   syncSourcei(master,slave,masterCtx,slaveCtx,AL_LOOPING);

 

   alcMakeContextCurrent(masterCtx);

   ALint source_type;

   alGetSourcei(master, AL_SOURCE_TYPE, &source_type);

 

   // Only static sources are currently synchronized! 

   if (AL_STATIC == source_type){

     ALint master_buffer;

     ALint slave_buffer;

     alGetSourcei(master, AL_BUFFER, &master_buffer);

     alcMakeContextCurrent(slaveCtx);

     alGetSourcei(slave, AL_BUFFER, &slave_buffer);

     if (master_buffer != slave_buffer){

       alSourcei(slave, AL_BUFFER, master_buffer);

     }

   }

   

   // Synchronize the state of the two sources.

   alcMakeContextCurrent(masterCtx);

   ALint masterState;

   ALint slaveState;

 

   alGetSourcei(master, AL_SOURCE_STATE, &masterState);

   alcMakeContextCurrent(slaveCtx);

   alGetSourcei(slave, AL_SOURCE_STATE, &slaveState);

 

   if (masterState != slaveState){

     switch (masterState){

     case AL_INITIAL : alSourceRewind(slave); break;

     case AL_PLAYING : alSourcePlay(slave);   break;

     case AL_PAUSED  : alSourcePause(slave);  break;

     case AL_STOPPED : alSourceStop(slave);   break;

     }

   }

   // Restore whatever context was previously active.

   alcMakeContextCurrent(current);

 }

 #+end_src

 This function is long because it has to exhaustively go through all the

 possible state that a source can have and make sure that it is the

 same between the master and slave sources.  I'd like to take this

 moment to salute the [[http://connect.creativelabs.com/openal/Documentation/Forms/AllItems.aspx][=OpenAL= Reference Manual]], which provides a very

 good description of =OpenAL='s internals.

 

 ** Context Synchronization

 #+name: sync-contexts

 #+begin_src C

 void syncContexts(ALCcontext *master, ALCcontext *slave){

   /* If there aren't sufficient sources in slave to mirror 

      the sources in master, create them. */

   ALCcontext *current = alcGetCurrentContext();

 

   UIntMap *masterSourceMap = &(master->SourceMap);

   UIntMap *slaveSourceMap = &(slave->SourceMap);

   ALuint numMasterSources = masterSourceMap->size;

   ALuint numSlaveSources = slaveSourceMap->size;

 

   alcMakeContextCurrent(slave);

   if (numSlaveSources < numMasterSources){

     ALuint numMissingSources = numMasterSources - numSlaveSources;

     ALuint newSources[numMissingSources];

     alGenSources(numMissingSources, newSources);

   }

 

   /* Now, slave is guaranteed to have at least as many sources

      as master.  Sync each source from master to the corresponding

      source in slave. */

   int i;

   for(i = 0; i < masterSourceMap->size; i++){

     syncSources((ALsource*)masterSourceMap->array[i].value,

 		(ALsource*)slaveSourceMap->array[i].value,

 		master, slave);

   }

   alcMakeContextCurrent(current);

 }

 #+end_src

 

 Most of the hard work in Context Synchronization is done in

 =syncSources()=. The only thing that =syncContexts()= has to worry

 about is automatically creating new sources whenever a slave context

 does not have the same number of sources as the master context.

 

 ** Context Creation

 #+name: context-creation

 #+begin_src C

 static void addContext(ALCdevice *Device, ALCcontext *context){

   send_data *data = (send_data*)Device->ExtraData;

   // expand array if necessary

   if (data->numContexts >= data->maxContexts){

     ALuint newMaxContexts = data->maxContexts*2 + 1;

     data->contexts = realloc(data->contexts, newMaxContexts*sizeof(context_data));

     data->maxContexts = newMaxContexts;

   }

   // create context_data and add it to the main array

   context_data *ctxData;

   ctxData = (context_data*)calloc(1, sizeof(*ctxData));

   ctxData->renderBuffer = 

     malloc(BytesFromDevFmt(Device->FmtType) * 

 	   Device->NumChan * Device->UpdateSize);

   ctxData->ctx = context;

 

   data->contexts[data->numContexts] = ctxData;

   data->numContexts++;

 }

 #+end_src

 

 Here, the slave context is created, and it's data is stored in the

 device-wide =ExtraData= structure.  The =renderBuffer= that is created

 here is where the rendered sound samples for this slave context will

 eventually go.

 

 ** Context Switching

 #+name: context-switching

 #+begin_src C

 ////////////////////  Context Switching 

 

 /* A device brings along with it two pieces of state

  * which have to be swapped in and out with each context.

  */

 static void swapInContext(ALCdevice *Device, context_data *ctxData){

   memcpy(Device->ClickRemoval, ctxData->ClickRemoval, sizeof(ALfloat)*MAXCHANNELS);

   memcpy(Device->PendingClicks, ctxData->PendingClicks, sizeof(ALfloat)*MAXCHANNELS);

 }

 

 static void saveContext(ALCdevice *Device, context_data *ctxData){

   memcpy(ctxData->ClickRemoval, Device->ClickRemoval, sizeof(ALfloat)*MAXCHANNELS);

   memcpy(ctxData->PendingClicks, Device->PendingClicks, sizeof(ALfloat)*MAXCHANNELS);

 }  

 

 static ALCcontext **currentContext;

 static ALuint currentNumContext;

 

 /* By default, all contexts are rendered at once for each call to aluMixData.

  * This function uses the internals of the ALCdevice struct to temporally 

  * cause aluMixData to only render the chosen context.

  */

 static void limitContext(ALCdevice *Device, ALCcontext *ctx){

   currentContext  = Device->Contexts;

   currentNumContext  = Device->NumContexts;

   Device->Contexts = &ctx;

   Device->NumContexts = 1;

 }

 

 static void unLimitContext(ALCdevice *Device){

   Device->Contexts = currentContext;

   Device->NumContexts = currentNumContext;

 }

 #+end_src

 

 =OpenAL= normally renders all contexts in parallel, outputting the

 whole result to the buffer.  It does this by iterating over the

 Device->Contexts array and rendering each context to the buffer in

 turn.  By temporally setting Device->NumContexts to 1 and adjusting

 the Device's context list to put the desired context-to-be-rendered

 into position 0, we can get trick =OpenAL= into rendering each context

 separate from all the others.

 

 ** Main Device Loop

 #+name: main-loop

 #+begin_src C

 ////////////////////   Main Device Loop

 

 /* Establish the LWJGL context as the master context, which will

  * be synchronized to all the slave contexts

  */

 static void init(ALCdevice *Device){

   ALCcontext *masterContext = alcGetCurrentContext();

   addContext(Device, masterContext);

 }

 

 static void renderData(ALCdevice *Device, int samples){

   if(!Device->Connected){return;}

   send_data *data = (send_data*)Device->ExtraData;

   ALCcontext *current = alcGetCurrentContext();

 

   ALuint i;

   for (i = 1; i < data->numContexts; i++){

     syncContexts(data->contexts[0]->ctx , data->contexts[i]->ctx);

   }

   

   if ((ALuint) samples > Device->UpdateSize){

     printf("exceeding internal buffer size; dropping samples\n");

     printf("requested %d; available %d\n", samples, Device->UpdateSize);

     samples = (int) Device->UpdateSize;

   }

 

   for (i = 0; i < data->numContexts; i++){

     context_data *ctxData = data->contexts[i];

     ALCcontext *ctx = ctxData->ctx;

     alcMakeContextCurrent(ctx);

     limitContext(Device, ctx);

     swapInContext(Device, ctxData);

     aluMixData(Device, ctxData->renderBuffer, samples);

     saveContext(Device, ctxData);

     unLimitContext(Device);

   }

   alcMakeContextCurrent(current);

 }

 #+end_src

 

 The main loop synchronizes the master LWJGL context with all the slave

 contexts, then iterates through each context, rendering just that

 context to it's audio-sample storage buffer.

 

 ** JNI Methods

 

 At this point, we have the ability to create multiple listeners by

 using the master/slave context trick, and the rendered audio data is

 waiting patiently in internal buffers, one for each listener.  We need

 a way to transport this information to Java, and also a way to drive

 this device from Java.  The following JNI interface code is inspired

 by the LWJGL JNI interface to =OpenAL=.

 

 *** Stepping the Device

 #+name: jni-step

 #+begin_src C

 ////////////////////   JNI Methods

 

 #include "com_aurellem_send_AudioSend.h"

 

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    nstep

  * Signature: (JI)V

  */

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_nstep

 (JNIEnv *env, jclass clazz, jlong device, jint samples){

   UNUSED(env);UNUSED(clazz);UNUSED(device);

   renderData((ALCdevice*)((intptr_t)device), samples);

 }

 #+end_src

 This device, unlike most of the other devices in =OpenAL=, does not

 render sound unless asked. This enables the system to slow down or

 speed up depending on the needs of the AIs who are using it to

 listen.  If the device tried to render samples in real-time, a

 complicated AI whose mind takes 100 seconds of computer time to

 simulate 1 second of AI-time would miss almost all of the sound in

 its environment.

 

 

 *** Device->Java Data Transport

 #+name: jni-get-samples

 #+begin_src C

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    ngetSamples

  * Signature: (JLjava/nio/ByteBuffer;III)V

  */

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_ngetSamples

 (JNIEnv *env, jclass clazz, jlong device, jobject buffer, jint position, 

  jint samples, jint n){

   UNUSED(clazz);  

 

   ALvoid *buffer_address = 

     ((ALbyte *)(((char*)(*env)->GetDirectBufferAddress(env, buffer)) + position));

   ALCdevice *recorder = (ALCdevice*) ((intptr_t)device);

   send_data *data = (send_data*)recorder->ExtraData;

   if ((ALuint)n > data->numContexts){return;}

   memcpy(buffer_address, data->contexts[n]->renderBuffer, 

 	 BytesFromDevFmt(recorder->FmtType) * recorder->NumChan * samples);

 }

 #+end_src

 

 This is the transport layer between C and Java that will eventually

 allow us to access rendered sound data from clojure.

 

 *** Listener Management

 

 =addListener=, =setNthListenerf=, and =setNthListener3f= are

 necessary to change the properties of any listener other than the

 master one, since only the listener of the current active context is

 affected by the normal =OpenAL= listener calls.

 #+name: listener-manage

 #+begin_src C

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    naddListener

  * Signature: (J)V

  */

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_naddListener

 (JNIEnv *env, jclass clazz, jlong device){

   UNUSED(env); UNUSED(clazz);

   //printf("creating new context via naddListener\n");

   ALCdevice *Device = (ALCdevice*) ((intptr_t)device);

   ALCcontext *new = alcCreateContext(Device, NULL);

   addContext(Device, new);

 }

 

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    nsetNthListener3f

  * Signature: (IFFFJI)V

  */

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_nsetNthListener3f

   (JNIEnv *env, jclass clazz, jint param, 

    jfloat v1, jfloat v2, jfloat v3, jlong device, jint contextNum){

   UNUSED(env);UNUSED(clazz);

 

   ALCdevice *Device = (ALCdevice*) ((intptr_t)device);

   send_data *data = (send_data*)Device->ExtraData;

   

   ALCcontext *current = alcGetCurrentContext();

   if ((ALuint)contextNum > data->numContexts){return;}

   alcMakeContextCurrent(data->contexts[contextNum]->ctx);

   alListener3f(param, v1, v2, v3);

   alcMakeContextCurrent(current);

 }

 

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    nsetNthListenerf

  * Signature: (IFJI)V

  */

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_nsetNthListenerf

 (JNIEnv *env, jclass clazz, jint param, jfloat v1, jlong device, 

  jint contextNum){

 

   UNUSED(env);UNUSED(clazz);

   

   ALCdevice *Device = (ALCdevice*) ((intptr_t)device);

   send_data *data = (send_data*)Device->ExtraData;

   

   ALCcontext *current = alcGetCurrentContext();

   if ((ALuint)contextNum > data->numContexts){return;}

   alcMakeContextCurrent(data->contexts[contextNum]->ctx);

   alListenerf(param, v1);

   alcMakeContextCurrent(current);

 }

 #+end_src

 

 *** Initialization

 =initDevice= is called from the Java side after LWJGL has created its

 context, and before any calls to =addListener=. It establishes the

 LWJGL context as the master context.

 

 =getAudioFormat= is a convenience function that uses JNI to build up a

 =javax.sound.sampled.AudioFormat= object from data in the Device. This

 way, there is no ambiguity about what the bits created by =step= and

 returned by =getSamples= mean.

 #+name: jni-init

 #+begin_src C

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    ninitDevice

  * Signature: (J)V

  */                     

 JNIEXPORT void JNICALL Java_com_aurellem_send_AudioSend_ninitDevice

 (JNIEnv *env, jclass clazz, jlong device){

   UNUSED(env);UNUSED(clazz);

   ALCdevice *Device = (ALCdevice*) ((intptr_t)device);

   init(Device);

 }

 

 /*

  * Class:     com_aurellem_send_AudioSend

  * Method:    ngetAudioFormat

  * Signature: (J)Ljavax/sound/sampled/AudioFormat;

  */

 JNIEXPORT jobject JNICALL Java_com_aurellem_send_AudioSend_ngetAudioFormat

 (JNIEnv *env, jclass clazz, jlong device){

   UNUSED(clazz);

   jclass AudioFormatClass = 

     (*env)->FindClass(env, "javax/sound/sampled/AudioFormat");

   jmethodID AudioFormatConstructor = 

     (*env)->GetMethodID(env, AudioFormatClass, "<init>", "(FIIZZ)V");

   

   ALCdevice *Device = (ALCdevice*) ((intptr_t)device);

   int isSigned;

   switch (Device->FmtType)

     {

     case DevFmtUByte: 

     case DevFmtUShort: isSigned = 0; break;  

     default : isSigned = 1;   

     }

   float frequency = Device->Frequency;

   int bitsPerFrame = (8 * BytesFromDevFmt(Device->FmtType));

   int channels = Device->NumChan;

   jobject format = (*env)->

     NewObject(

 	      env,AudioFormatClass,AudioFormatConstructor,

 	      frequency,

 	      bitsPerFrame,

 	      channels,

 	      isSigned,

 	      0);

   return format;

 }

 #+end_src

 

 ** Boring Device Management Stuff / Memory Cleanup

 This code is more-or-less copied verbatim from the other =OpenAL=

 Devices. It's the basis for =OpenAL='s primitive object system.

 #+name: device-init

 #+begin_src C

 ////////////////////   Device Initialization / Management

 

 static const ALCchar sendDevice[] = "Multiple Audio Send";

 

 static ALCboolean send_open_playback(ALCdevice *device, 

  const ALCchar *deviceName)

 {

   send_data *data;

   // stop any buffering for stdout, so that I can 

   // see the printf statements in my terminal immediately

   setbuf(stdout, NULL);

 

   if(!deviceName)

     deviceName = sendDevice;

   else if(strcmp(deviceName, sendDevice) != 0)

     return ALC_FALSE;

   data = (send_data*)calloc(1, sizeof(*data));

   device->szDeviceName = strdup(deviceName);

   device->ExtraData = data;

   return ALC_TRUE;

 }

 

 static void send_close_playback(ALCdevice *device)

 {

   send_data *data = (send_data*)device->ExtraData;

   alcMakeContextCurrent(NULL);

   ALuint i;

   // Destroy all slave contexts. LWJGL will take care of 

   // its own context.

   for (i = 1; i < data->numContexts; i++){

     context_data *ctxData = data->contexts[i];

     alcDestroyContext(ctxData->ctx);

     free(ctxData->renderBuffer);

     free(ctxData);

   }

   free(data);

   device->ExtraData = NULL;

 }

 

 static ALCboolean send_reset_playback(ALCdevice *device)

 {

   SetDefaultWFXChannelOrder(device);

   return ALC_TRUE;

 }

 

 static void send_stop_playback(ALCdevice *Device){

   UNUSED(Device);

 }

 

 static const BackendFuncs send_funcs = {

   send_open_playback,

   send_close_playback,

   send_reset_playback,

   send_stop_playback,

   NULL,

   NULL,  /* These would be filled with functions to    */

   NULL,  /* handle capturing audio if we we into that  */

   NULL,  /* sort of thing...                           */

   NULL,

   NULL

 };

 

 ALCboolean alc_send_init(BackendFuncs *func_list){

   *func_list = send_funcs;

   return ALC_TRUE;

 }

 

 void alc_send_deinit(void){}

 

 void alc_send_probe(enum DevProbe type)

 {

   switch(type)

     {

     case DEVICE_PROBE:

       AppendDeviceList(sendDevice);

       break;

     case ALL_DEVICE_PROBE:

       AppendAllDeviceList(sendDevice);

       break;

     case CAPTURE_DEVICE_PROBE:

       break;

     }

 }

 #+end_src

 

 * The Java interface, =AudioSend=

 

 The Java interface to the Send Device follows naturally from the JNI

 definitions. The only thing here of note is the =deviceID=. This is

 available from LWJGL, but to only way to get it is with reflection.

 Unfortunately, there is no other way to control the Send device than

 to obtain a pointer to it.

 

 #+include: "../../audio-send/java/src/com/aurellem/send/AudioSend.java" src java 

 

 * The Java Audio Renderer, =AudioSendRenderer=

 

 #+include: "../../jmeCapture/src/com/aurellem/capture/audio/AudioSendRenderer.java" src java

 

 The =AudioSendRenderer= is a modified version of the

 =LwjglAudioRenderer= which implements the =MultiListener= interface to

 provide access and creation of more than one =Listener= object.

 

 ** MultiListener.java

 

 #+include: "../../jmeCapture/src/com/aurellem/capture/audio/MultiListener.java" src java

 

 ** SoundProcessors are like SceneProcessors

 

 A =SoundProcessor= is analgous to a =SceneProcessor=. Every frame, the

 =SoundProcessor= registered with a given =Listener= recieves the

 rendered sound data and can do whatever processing it wants with it.

 

 #+include "../../jmeCapture/src/com/aurellem/capture/audio/SoundProcessor.java" src java  

 

 * Finally, Ears in clojure! 

 

 Now that the =C= and =Java= infrastructure is complete, the clojure

 hearing abstraction is simple and closely parallels the [[./vision.org][vision]]

 abstraction.

 

 ** Hearing Pipeline

 

 All sound rendering is done in the CPU, so =(hearing-pipeline)= is

 much less complicated than =(vision-pipelie)= The bytes available in

 the ByteBuffer obtained from the =send= Device have different meanings

 dependant upon the particular hardware or your system.  That is why

 the =AudioFormat= object is necessary to provide the meaning that the

 raw bytes lack. =(byteBuffer->pulse-vector)= uses the excellent

 conversion facilities from [[http://www.tritonus.org/ ][tritonus]] ([[http://tritonus.sourceforge.net/apidoc/org/tritonus/share/sampled/FloatSampleTools.html#byte2floatInterleaved%2528byte%5B%5D,%2520int,%2520float%5B%5D,%2520int,%2520int,%2520javax.sound.sampled.AudioFormat%2529][javadoc]]) to generate a clojure vector of

 floats which represent the linear PCM encoded waveform of the

 sound. With linear PCM (pulse code modulation) -1.0 represents maximum

 rarefaction of the air while 1.0 represents maximum compression of the

 air at a given instant.

 

 #+name: hearing-pipeline

 #+begin_src clojure

 (in-ns 'cortex.hearing)

 

 (defn hearing-pipeline

   "Creates a SoundProcessor which wraps a sound processing

   continuation function. The continuation is a function that takes

   [#^ByteBuffer b #^Integer int numSamples #^AudioFormat af ], each of which

   has already been apprpiately sized."

   [continuation]

   (proxy [SoundProcessor] []

     (cleanup [])

     (process

       [#^ByteBuffer audioSamples numSamples #^AudioFormat audioFormat]

       (continuation audioSamples numSamples audioFormat))))

 

 (defn byteBuffer->pulse-vector

   "Extract the sound samples from the byteBuffer as a PCM encoded

    waveform with values ranging from -1.0 to 1.0 into a vector of

    floats." 

   [#^ByteBuffer audioSamples numSamples #^AudioFormat audioFormat]

   (let [num-floats (/ numSamples  (.getFrameSize audioFormat))

         bytes  (byte-array numSamples)

         floats (float-array num-floats)]

     (.get audioSamples bytes 0 numSamples)

     (FloatSampleTools/byte2floatInterleaved

      bytes 0 floats 0 num-floats audioFormat)

     (vec floats)))

 #+end_src

 

 ** Physical Ears

 

 Together, these three functions define how ears found in a specially

 prepared blender file will be translated to =Listener= objects in a

 simulation. =(ears)= extracts all the children of to top level node

 named "ears".  =(add-ear!)= and =(update-listener-velocity!)= use

 =(bind-sense)= to bind a =Listener= object located at the initial

 position of an "ear" node to the closest physical object in the

 creature. That =Listener= will stay in the same orientation to the

 object with which it is bound, just as the camera in the [[http://aurellem.localhost/cortex/html/sense.html#sec-4-1][sense binding

 demonstration]].  =OpenAL= simulates the doppler effect for moving

 listeners, =(update-listener-velocity!)= ensures that this velocity

 information is always up-to-date.

 

 #+name: hearing-ears

 #+begin_src clojure

 (defvar 

   ^{:arglists '([creature])}

   ears

   (sense-nodes "ears")

   "Return the children of the creature's \"ears\" node.")

 

 (defn update-listener-velocity!

   "Update the listener's velocity every update loop."

   [#^Spatial obj #^Listener lis]

   (let [old-position (atom (.getLocation lis))]

     (.addControl

      obj

      (proxy [AbstractControl] []

        (controlUpdate [tpf]

          (let [new-position (.getLocation lis)]

            (.setVelocity

             lis 

             (.mult (.subtract new-position @old-position)

                    (float (/ tpf))))

            (reset! old-position new-position)))

        (controlRender [_ _])))))

 

 (defn add-ear!  

   "Create a Listener centered on the current position of 'ear 

    which follows the closest physical node in 'creature and 

    sends sound data to 'continuation."

   [#^Application world #^Node creature #^Spatial ear continuation]

   (let [target (closest-node creature ear)

         lis (Listener.)

         audio-renderer (.getAudioRenderer world)

         sp (hearing-pipeline continuation)]

     (.setLocation lis (.getWorldTranslation ear))

     (.setRotation lis (.getWorldRotation ear))

     (bind-sense target lis)

     (update-listener-velocity! target lis)

     (.addListener audio-renderer lis)

     (.registerSoundProcessor audio-renderer lis sp)))

 #+end_src

 

 ** Ear Creation

 

 #+name: hearing-kernel

 #+begin_src clojure

 (defn hearing-kernel

   "Returns a functon which returns auditory sensory data when called

    inside a running simulation."

   [#^Node creature #^Spatial ear]

   (let [hearing-data (atom [])

         register-listener!

         (runonce 

          (fn [#^Application world]

            (add-ear!

             world creature ear

             (comp #(reset! hearing-data %)

                   byteBuffer->pulse-vector))))]

     (fn [#^Application world]

       (register-listener! world)

       (let [data @hearing-data

             topology              

             (vec (map #(vector % 0) (range 0 (count data))))]

         [topology data]))))

     

 (defn hearing!

   "Endow the creature in a particular world with the sense of

    hearing. Will return a sequence of functions, one for each ear,

    which when called will return the auditory data from that ear."

   [#^Node creature]

   (for [ear (ears creature)]

     (hearing-kernel creature ear)))

 #+end_src

 

 Each function returned by =(hearing-kernel!)= will register a new

 =Listener= with the simulation the first time it is called.  Each time

 it is called, the hearing-function will return a vector of linear PCM

 encoded sound data that was heard since the last frame. The size of

 this vector is of course determined by the overall framerate of the

 game. With a constant framerate of 60 frames per second and a sampling

 frequency of 44,100 samples per second, the vector will have exactly

 735 elements.

 

 ** Visualizing Hearing

 

 This is a simple visualization function which displaye the waveform

 reported by the simulated sense of hearing. It converts the values

 reported in the vector returned by the hearing function from the range

 [-1.0, 1.0] to the range [0 255], converts to integer, and displays

 the number as a greyscale pixel.

 

 #+name: hearing-display

 #+begin_src clojure

 (in-ns 'cortex.hearing)

 

 (defn view-hearing

   "Creates a function which accepts a list of auditory data and

    display each element of the list to the screen as an image."

   []

   (view-sense

    (fn [[coords sensor-data]]

      (let [pixel-data 

            (vec

             (map

              #(rem (int (* 255 (/ (+ 1 %) 2))) 256)

              sensor-data))

            height 50

            image (BufferedImage. (max 1 (count coords)) height

                                  BufferedImage/TYPE_INT_RGB)]

        (dorun

         (for [x (range (count coords))]

           (dorun

            (for [y (range height)]

              (let [raw-sensor (pixel-data x)]

                (.setRGB image x y (gray raw-sensor)))))))

        image))))

 #+end_src

 

 * Testing Hearing

 

 ** Advanced Java Example

 

 I wrote a test case in Java that demonstrates the use of the Java

 components of this hearing system. It is part of a larger java library

 to capture perfect Audio from jMonkeyEngine. Some of the clojure

 constructs above are partially reiterated in the java source file. But

 first, the video! As far as I know this is the first instance of

 multiple simulated listeners in a virtual environment using OpenAL.

 

 #+begin_html

 <div class="figure">

 <center>

 <video controls="controls" width="500">

   <source src="../video/java-hearing-test.ogg" type="video/ogg"

 	  preload="none" poster="../images/aurellem-1280x480.png" />

 </video>

 </center>

 <p>The blue ball is emitting a constant sound. Each blue box is

   listening for sound, and will change color from blue to green if it

   detects sound which is louder than a certain threshold. As the blue

   sphere travels along the path, it excites each of the cubes in turn.</p>

 </div>

 #+end_html

 

 #+include "../../jmeCapture/src/com/aurellem/capture/examples/Advanced.java" src java  

 

 Here is a small clojure program to drive the java program and make it

 available as part of my test suite.

 

 #+name: test-hearing-1

 #+begin_src clojure

 (in-ns 'cortex.test.hearing)

 

 (defn test-java-hearing

   "Testing hearing:

    You should see a blue sphere flying around several

    cubes.  As the sphere approaches each cube, it turns

    green."

   []

   (doto (com.aurellem.capture.examples.Advanced.)

     (.setSettings

      (doto (AppSettings. true)

        (.setAudioRenderer "Send")))

     (.setShowSettings false)

     (.setPauseOnLostFocus false)))

 #+end_src

 

 

 

 

 ** Adding Hearing to the Worm

 

 To the worm, I add a new node called "ears" with one child which

 represents the worm's single ear.

 

 #+attr_html: width=755

 #+caption: The Worm with a newly added nodes describing an ear.

 [[../images/worm-with-ear.png]]

 

 The node highlighted in yellow it the top-level "ears" node. It's

 child, highlighted in orange, represents a the single ear the creature

 has. The ear will be localized right above the curved part of the

 worm's lower hemispherical region opposite the eye.  

 

 The other empty nodes represent the worm's single joint and eye and are

 described in [[./body.org][body]] and [[./vision.org][vision]].

 

 #+name: test-hearing-2

 #+begin_src clojure 

 (in-ns 'cortex.test.hearing)

 

 (cortex.import/mega-import-jme3)

 (import java.io.File)

 

 (use 'cortex.body)

 

 (defn test-worm-hearing []

   (let [the-worm (doto (worm) (body!))

         hearing (hearing! the-worm)

         hearing-display (view-hearing)

         

         tone (AudioNode. (asset-manager)

                          "Sounds/pure.wav" false)

         

         hymn (AudioNode. (asset-manager) 

                          "Sounds/ear-and-eye.wav" false)]

     (world

      (nodify [the-worm (floor)])

      (merge standard-debug-controls

             {"key-return"

              (fn [_ value]

                (if value (.play tone)))

              "key-l"

              (fn [_ value]

                (if value (.play hymn)))})

      (fn [world]

        (light-up-everything world)

        (com.aurellem.capture.Capture/captureVideo

         world

         (File."/home/r/proj/cortex/render/worm-audio/frames"))

        (com.aurellem.capture.Capture/captureAudio

         world

         (File."/home/r/proj/cortex/render/worm-audio/audio.wav")))

      

      (fn [world tpf]

        (hearing-display

         (map #(% world) hearing)

         (File. "/home/r/proj/cortex/render/worm-audio/hearing-data"))))))

 #+end_src

 

 In this test, I load the worm with its newly formed ear and let it

 hear sounds. The sound the worm is hearing is localized to the origin

 of the world, and you can see that as the worm moves farther away from

 the origin when it is hit by balls, it hears the sound less intensely.

 

 The sound you hear in the video is from the worm's perspective. Notice

 how the pure tone becomes fainter and the visual display of the

 auditory data becomes less pronounced as the worm falls farther away

 from the source of the sound.

 

 #+begin_html

 <div class="figure">

 <center>

 <video controls="controls" width="600">

   <source src="../video/worm-hearing.ogg" type="video/ogg"

 	  preload="none" poster="../images/aurellem-1280x480.png" />

 </video>

 </center>

 <p>The worm can now hear the sound pulses produced from the

   hymn. Notice the strikingly different pattern that human speech

   makes compared to the insturments. Once the worm is pushed off the

   floor, the sound it hears is attenuated, and the display of the

   sound it hears beomes fainter. This shows the 3D localization of

   sound in this world.</p>

 </div>

 

 #+end_html

 

 *** Creating the Ear Video

 #+name: magick-3

 #+begin_src clojure

 (ns cortex.video.magick3

   (:import java.io.File)

   (:use clojure.contrib.shell-out))

 

 (defn images [path]

   (sort (rest (file-seq (File. path)))))

 

 (def base "/home/r/proj/cortex/render/worm-audio/")

 

 (defn pics [file]

   (images (str base file)))

 

 (defn combine-images []

   (let [main-view (pics "frames")

         hearing (pics "hearing-data")

         background (repeat 9001 (File. (str base "background.png")))

         targets (map

                  #(File. (str base "out/" (format "%07d.png" %)))

                  (range 0 (count main-view)))]

     (dorun

      (pmap

       (comp

        (fn [[background main-view hearing target]]

          (println target)

          (sh "convert"

              background

              main-view "-geometry" "+66+21"   "-composite"

              hearing   "-geometry" "+21+526"  "-composite"

              target))

        (fn [& args] (map #(.getCanonicalPath %) args)))

       background main-view hearing targets))))

 #+end_src

 

 #+begin_src sh

 cd /home/r/proj/cortex/render/worm-audio

 ffmpeg -r 60 -i out/%07d.png -i audio.wav \

        -b:a 128k  -b:v 9001k \

        -acodec libvorbis -vcodec libtheora worm-hearing.ogg 

 #+end_src

 

 * Headers

 

 #+name: hearing-header

 #+begin_src clojure

 (ns cortex.hearing

   "Simulate the sense of hearing in jMonkeyEngine3. Enables multiple

   listeners at different positions in the same world. Automatically

   reads ear-nodes from specially prepared blender files and

   instantiates them in the world as simulated ears."

   {:author "Robert McIntyre"}

   (:use (cortex world util sense))

   (:use clojure.contrib.def)

   (:import java.nio.ByteBuffer)

   (:import java.awt.image.BufferedImage)

   (:import org.tritonus.share.sampled.FloatSampleTools)

   (:import (com.aurellem.capture.audio

             SoundProcessor AudioSendRenderer))

   (:import javax.sound.sampled.AudioFormat)

   (:import (com.jme3.scene Spatial Node))

   (:import com.jme3.audio.Listener)

   (:import com.jme3.app.Application)

   (:import com.jme3.scene.control.AbstractControl))

 #+end_src

 

 #+name: test-header

 #+begin_src clojure

 (ns cortex.test.hearing

   (:use (cortex world util hearing))

   (:use cortex.test.body)

   (:import (com.jme3.audio AudioNode Listener))

   (:import com.jme3.scene.Node

 	   com.jme3.system.AppSettings))

 #+end_src

 

 

 * Next

  - As a bonus, this method of capturing audio for AI can also be used

    to capture perfect audio from a jMonkeyEngine application, for use

    in demos and the like.

 

 

 

 * COMMENT Code Generation

 

 #+begin_src clojure :tangle ../src/cortex/hearing.clj

 <<hearing-header>>

 <<hearing-pipeline>>

 <<hearing-ears>>

 <<hearing-kernel>>

 <<hearing-display>>

 #+end_src

 

 #+begin_src clojure :tangle ../src/cortex/test/hearing.clj

 <<test-header>>

 <<test-hearing-1>>

 <<test-hearing-2>>

 #+end_src

 

 #+begin_src clojure :tangle ../src/cortex/video/magick3.clj

 <<magick-3>>

 #+end_src

 

 #+begin_src C :tangle ../../audio-send/Alc/backends/send.c

 <<send-header>>

 <<send-state>>

 <<sync-macros>>

 <<sync-sources>>

 <<sync-contexts>>

 <<context-creation>>

 <<context-switching>>

 <<main-loop>>

 <<jni-step>>

 <<jni-get-samples>>

 <<listener-manage>>

 <<jni-init>>

 <<device-init>>

 #+end_src